Currently, the plan is being reviewed by President Trump's chief of staff, Susie Wiles. President Trump hasn't examined the plan yet, meaning the specifics might still shift. Unclear details include the types of model evaluations the plan would require, the funding mechanism for a new agency, and the SEC's involvement.
Venture capitalist David Sacks, who previously served as White House AI czar and still advises Trump, said the new Chinese model was "concerning" and that it stemmed from the U.S. "tying itself in knots" through mandatory governmental sign‑offs on AI systems.
Last month, U.S. export restrictions compelled Anthropic PBC, a private AI firm, to temporarily take down its Fable 5 and Mythos 5 systems because of U.S. export restrictions. Then on Friday, Chinese startup Moonshot AI Inc. unveiled its Kimi K3 model, causing investors to question whether American companies can sustain their large expenditures on semiconductor hardware and data centers.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
The Kimi K3 model demonstrated significant performance gains, triggering a selloff in AI-related stocks as fears mounted that U.S. firms might lose their competitive edge. The episode underscored the need for a more predictable regulatory environment, especially as Chinese AI capabilities rapidly advance.
Industry leaders are already lining up behind the idea. Earlier this week, Google DeepMind CEO Demis Hassabis released policy recommendations that propose a panel of independent technical experts to evaluate AI models within a standards organization supervised by the federal government. Microsoft's Satya Nadella, along with OpenAI's Sam Altman and Elon Musk of SpaceX, all expressed support for the proposals.
Hassabis has a meeting set with Washington policymakers next week to advocate for his proposal. In a post on X, Hassabis likened the watchdog to Finra.
Currently, the government has the ability to request modifications to models - for instance, it did so with OpenAI's Sol model - or prevent their distribution via export restrictions. This kind of framework would give greater assurance to leading AI labs such as Anthropic and OpenAI, which both opposed the government's previous actions.
The catch: negotiations could drag on, and the final version might look different from what is on the table now. Both OpenAI and Anthropic are still private firms, but each is considering an IPO in the coming year.
The Trump administration is trying to balance two competing needs. Wall Street firms want to minimize AI's cybersecurity risks. Tech firms in Silicon Valley are annoyed by what they perceive as a lack of consistency in the administration's approach. The proposal aims to appease both sides.
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets